source: sasview/pr_inversion/test/utest_invertor.py @ 383189f9

"""
    Unit tests for Invertor class
"""
# Disable "missing docstring" complaint
# pylint: disable-msg=C0111
# Disable "too many methods" complaint 
# pylint: disable-msg=R0904 


import unittest, math, numpy, pylab
from sans.pr.invertor import Invertor
        
class TestBasicComponent(unittest.TestCase):
    
    def setUp(self):
        self.invertor = Invertor()
        self.invertor.d_max = 100.0
        
        # Test array
        self.ntest = 5
        self.x_in = numpy.ones(self.ntest)
        for i in range(self.ntest):
            self.x_in[i] = 1.0*(i+1)


    def testset_dmax(self):
        """
            Set and read d_max
        """
        value = 15.0
        self.invertor.d_max = value
        self.assertEqual(self.invertor.d_max, value)
        
    def testset_alpha(self):
        """
            Set and read alpha
        """
        value = 15.0
        self.invertor.alpha = value
        self.assertEqual(self.invertor.alpha, value)
        
    def testset_x_1(self):
        """
            Setting and reading the x array the hard way
        """
        # Set x
        self.invertor.x = self.x_in
        
        # Read it back
        npts = self.invertor.get_nx()
        x_out = numpy.ones(npts)
        
        self.invertor.get_x(x_out)

        for i in range(self.ntest):
            self.assertEqual(self.x_in[i], x_out[i])
            
    def testset_x_2(self):
        """
            Setting and reading the x array the easy way
        """
        # Set x
        self.invertor.x = self.x_in
        
        # Read it back
        x_out = self.invertor.x
        
        for i in range(self.ntest):
            self.assertEqual(self.x_in[i], x_out[i])
       
    def testset_y(self):
        """
            Setting and reading the y array the easy way
        """
        # Set y
        self.invertor.y = self.x_in
        
        # Read it back
        y_out = self.invertor.y
        
        for i in range(self.ntest):
            self.assertEqual(self.x_in[i], y_out[i])
       
    def testset_err(self):
        """
            Setting and reading the err array the easy way
        """
        # Set err
        self.invertor.err = self.x_in
        
        # Read it back
        err_out = self.invertor.err
        
        for i in range(self.ntest):
            self.assertEqual(self.x_in[i], err_out[i])
       
    def test_iq(self):
        """
            Test iq calculation
        """
        q = 0.11
        v1 = 8.0*math.pi**2/q * self.invertor.d_max *math.sin(q*self.invertor.d_max)
        v1 /= ( math.pi**2 - (q*self.invertor.d_max)**2.0 )
        
        pars = numpy.ones(1)
        self.assertAlmostEqual(self.invertor.iq(pars, q), v1, 2)
        
    def test_pr(self):
        """
            Test pr calculation
        """
        r = 10.0
        v1 = 2.0*r*math.sin(math.pi*r/self.invertor.d_max)
        pars = numpy.ones(1)
        self.assertAlmostEqual(self.invertor.pr(pars, r), v1, 2)
        
    def test_getsetters(self):
        self.invertor.new_data = 1.0
        self.assertEqual(self.invertor.new_data, 1.0)
        
        self.assertEqual(self.invertor.test_no_data, None)
        
    def test_inversion(self):
        """
            Test an inversion for which we know the answer
        """
        x, y, err = load("sphere_80.txt")

        # Choose the right d_max...
        self.invertor.d_max = 160.0
        # Set a small alpha
        self.invertor.alpha = 1e-7
        # Set data
        self.invertor.x   = x
        self.invertor.y   = y
        self.invertor.err = err
        # Perform inversion
        out, cov = self.invertor.invert(10)
        # This is a very specific case
        # We should make sure it always passes
        self.assertTrue(self.invertor.chi2/len(x)<200.00)
        
        # Check the computed P(r) with the theory
        # for shpere of radius 80
        x = pylab.arange(0.01, self.invertor.d_max, self.invertor.d_max/51.0)
        y = numpy.zeros(len(x))
        dy = numpy.zeros(len(x))
        y_true = numpy.zeros(len(x))

        sum = 0.0
        sum_true = 0.0
        for i in range(len(x)):
            #y[i] = self.invertor.pr(out, x[i])
            (y[i], dy[i]) = self.invertor.pr_err(out, cov, x[i])
            sum += y[i]
            if x[i]<80.0:
                y_true[i] = pr_theory(x[i], 80.0)
            else:
                y_true[i] = 0
            sum_true += y_true[i]
            
        y = y/sum*self.invertor.d_max/len(x)
        dy = dy/sum*self.invertor.d_max/len(x)
        y_true = y_true/sum_true*self.invertor.d_max/len(x)
        
        chi2 = 0.0
        for i in range(len(x)):
            res = (y[i]-y_true[i])/dy[i]
            chi2 += res*res
            
        try:
            self.assertTrue(chi2/51.0<10.0)
        except:
            print "chi2 =", chi2/51.0
            raise
        
    def test_lstsq(self):
        """
            Test an inversion for which we know the answer
        """
        x, y, err = load("sphere_80.txt")

        # Choose the right d_max...
        self.invertor.d_max = 160.0
        # Set a small alpha
        self.invertor.alpha = .0007
        # Set data
        self.invertor.x   = x
        self.invertor.y   = y
        self.invertor.err = err
        # Perform inversion
        #out, cov = self.invertor.invert(10)
        
        out, cov = self.invertor.lstsq(10)
         
        
        # This is a very specific case
        # We should make sure it always passes
        try:
            self.assertTrue(self.invertor.chi2/len(x)<200.00)
        except:
            print "Chi2(I(q)) =", self.invertor.chi2/len(x)
            raise
        
        # Check the computed P(r) with the theory
        # for shpere of radius 80
        x = pylab.arange(0.01, self.invertor.d_max, self.invertor.d_max/51.0)
        y = numpy.zeros(len(x))
        dy = numpy.zeros(len(x))
        y_true = numpy.zeros(len(x))

        sum = 0.0
        sum_true = 0.0
        for i in range(len(x)):
            #y[i] = self.invertor.pr(out, x[i])
            (y[i], dy[i]) = self.invertor.pr_err(out, cov, x[i])
            sum += y[i]
            if x[i]<80.0:
                y_true[i] = pr_theory(x[i], 80.0)
            else:
                y_true[i] = 0
            sum_true += y_true[i]
            
        y = y/sum*self.invertor.d_max/len(x)
        dy = dy/sum*self.invertor.d_max/len(x)
        y_true = y_true/sum_true*self.invertor.d_max/len(x)
        
        chi2 = 0.0
        for i in range(len(x)):
            res = (y[i]-y_true[i])/dy[i]
            chi2 += res*res
            
        try:
            self.assertTrue(chi2/51.0<50.0)
        except:
            print "chi2(P(r)) =", chi2/51.0
            raise
            
    def test_q_zero(self):
        """
            Test error condition where a point has q=0
        """
        x, y, err = load("sphere_80.txt")
        x[0] = 0.0
        
        # Choose the right d_max...
        self.invertor.d_max = 160.0
        # Set a small alpha
        self.invertor.alpha = 1e-7
        # Set data
        def doit():
            self.invertor.x   = x
        self.assertRaises(ValueError, doit)
        
                            
    def test_q_neg(self):
        """
            Test error condition where a point has q<0
        """
        x, y, err = load("sphere_80.txt")
        x[0] = -0.2
        
        # Choose the right d_max...
        self.invertor.d_max = 160.0
        # Set a small alpha
        self.invertor.alpha = 1e-7
        # Set data
        self.invertor.x   = x
        self.invertor.y   = y
        self.invertor.err = err
        # Perform inversion
        out, cov = self.invertor.invert(4)
        
        try:
            self.assertTrue(self.invertor.chi2>0)
        except:
            print "Chi2 =", self.invertor.chi2
            raise
                            
    def test_Iq_zero(self):
        """
            Test error condition where a point has q<0
        """
        x, y, err = load("sphere_80.txt")
        y[0] = 0.0
        
        # Choose the right d_max...
        self.invertor.d_max = 160.0
        # Set a small alpha
        self.invertor.alpha = 1e-7
        # Set data
        self.invertor.x   = x
        self.invertor.y   = y
        self.invertor.err = err
        # Perform inversion
        out, cov = self.invertor.invert(4)
        
        try:
            self.assertTrue(self.invertor.chi2>0)
        except:
            print "Chi2 =", self.invertor.chi2
            raise
        
    def no_test_time(self):
        x, y, err = load("sphere_80.txt")

        # Choose the right d_max...
        self.invertor.d_max = 160.0
        # Set a small alpha
        self.invertor.alpha = 1e-7
        # Set data
        self.invertor.x   = x
        self.invertor.y   = y
        self.invertor.err = err
    
        # time scales like nfunc**2
        # on a Lenovo Intel Core 2 CPU T7400 @ 2.16GHz, 
        # I get time/(nfunc)**2 = 0.022 sec
                            
        out, cov = self.invertor.invert(15)
        t16 = self.invertor.elapsed
        
        out, cov = self.invertor.invert(30)
        t30 = self.invertor.elapsed
        
        t30s = t30/30.0**2
        self.assertTrue( (t30s-t16/16.0**2)/t30s <1.2 )
        
    def test_clone(self):
        self.invertor.x = self.x_in
        clone = self.invertor.clone()
        
        for i in range(len(self.x_in)):
            self.assertEqual(self.x_in[i], clone.x[i])
        
def pr_theory(r, R):
    """
       P(r) for a sphere
    """
    if r<=2*R:
        return 12.0* ((0.5*r/R)**2) * ((1.0-0.5*r/R)**2) * ( 2.0 + 0.5*r/R )
    else:
        return 0.0
    
def load(path = "sphere_60_q0_2.txt"):
    import numpy, math, sys
    # Read the data from the data file
    data_x   = numpy.zeros(0)
    data_y   = numpy.zeros(0)
    data_err = numpy.zeros(0)
    if not path == None:
        input_f = open(path,'r')
        buff    = input_f.read()
        lines   = buff.split('\n')
        for line in lines:
            try:
                toks = line.split()
                x = float(toks[0])
                y = float(toks[1])
                data_x = numpy.append(data_x, x)
                data_y = numpy.append(data_y, y)
                # Set the error of the first point to 5%
                # to make theory look like data
                scale = 0.1/math.sqrt(data_x[0])
                data_err = numpy.append(data_err, scale*math.sqrt(y))
            except:
                pass
               
    return data_x, data_y, data_err 
       
if __name__ == '__main__':
    unittest.main()